Work / Lab
AI due-diligence market map
Teardown of the AI due-diligence market: six segments, feature-by-feature defensibility, and what actually survives.
- Claude Code (research workflow)
TL;DR
- After running Concord live, I asked the obvious question: does this already exist? I directed a multi-agent research sweep to find out.
- It split the field into six segments and scored every Concord feature against them.
- Most features came back commoditised. Two survived, and the top recommendation became the next thing I built.
The sweep
The segments: PE-diligence software, legal AI moving into diligence, data-room incumbents, Big 4 advisory, frontier labs and hyperscalers, and open source. For each Concord feature, the sweep asked who already ships it and how hard it would be to copy.
What came back
- Commoditised: multi-model routing, bring-your-own-key desktop apps, house playbooks and devil’s-advocate review.
- Survived: a blocking tie-out that checks every cited quote against the source document and produces a per-run score, and per-finding cross-vendor challenge that preserves dissent instead of averaging it away.
- Top recommendation: a public adversarial benchmark. I built it next; it isn’t published yet.
Honest notes
The sweep ran after Concord was built, not before, and it recommended repositioning around the two surviving pieces rather than stopping. The question and the follow-through were mine; the research itself was agent-generated at my request.